What SpaceX’s First-Ever Earnings Reveal About Its Business

• SpaceX’s debut quarter showed one excellent business bankrolling an enormous AI.
• However, the profit still remains an accounting artifact.
• Two days later, a share unlock larger than the entire public float begins.

SpaceX () reported for the first time on Tuesday, and the timing is hard to ignore. The numbers crossed the wire after the close on August 4, the first audited look at a company that listed only in June.

Source: InvestingPro

Two days later, on August 6, the first tranche of insider stock unlocks. Because SpaceX floated only about 4.9% of itself at the IPO, after Nasdaq waived its usual 10 percent minimum, that first unlock alone frees more shares than the entire public float. The debut quarter and the first real test of supply arrive in the same week.

That collision is the frame for everything that follows. A listing this tightly held was always as much a staged liquidity event as a capital raise, and for the next several months the share price will be driven as much by who becomes free to sell as by how the company performs. Which is precisely why the more useful exercise is to look past the tape and ask whether the business underneath is actually working.

On that question, the first print is more encouraging than the headline loss suggests, and more complicated than the headline profit suggests. SpaceX generated $3.5 billion of adjusted EBITDA in the quarter and spent $18.4 billion building the business over the same three months. Those two figures, sitting side by side, are the whole story.

How Should the Quarter Be Read?

The clearest way to make sense of the numbers is to treat SpaceX as one excellent business funding two very large bets. Starlink is the excellent business. Starship and the artificial intelligence compute build are the bets. Almost every figure that looks confusing at first resolves once those three are separated.

The listing itself supplied the capital for those bets. The June IPO was an all-primary offering, meaning no existing holder sold, that placed 638.9 million Class A shares at $135 and raised roughly $85.7 billion of net proceeds at a valuation near $1.77 trillion, the largest IPO on record, with about 30 percent allocated to retail. The money on the balance sheet is therefore primary capital rather than the proceeds of insiders cashing out, and the public float is unusually thin.

On a consolidated basis, revenue reached $7.8 billion, up 92 percent from a year earlier and about 67 percent from the prior quarter. Adjusted EBITDA came in at $3.5 billion, and the operating result was close to breakeven at a loss of $143 million, against a $970 million operating loss a year ago.

The company still posted a net loss of $541 million, or nine cents a share, though that was roughly half the loss of a year earlier. The important caveat, developed below, is that the adjusted profitability rests heavily on add-backs, chiefly depreciation, so the strong EBITDA line and the near-breakeven operating line describe a company investing far more than it currently earns.

Source: SpaceX Q2 2026 results.

Revenue nearly doubled year over year, and the step up in the June quarter is almost entirely the AI segment. Connectivity remains the base of the business.

Source: SpaceX Q2 2026 results.

How Healthy Is the Core at Starlink?

Connectivity is where the company actually makes money, and it strengthened. Segment revenue rose 66 percent year over year to $4.3 billion, income from operations grew 79 percent to $1.7 billion, and segment adjusted EBITDA reached $2.6 billion. This is not a new development. Already in the prospectus, Connectivity’s EBITDA exceeded the entire company’s EBITDA, which means every other segment was net negative and the core was carrying the group even before the listing.

The question going into the print was pricing. Blended ARPU had fallen from $85 a year earlier toward the mid sixties, and the worry was that it would keep sliding. Instead ARPU held flat sequentially at $66 while subscribers reached 12.0 million, double a year earlier and up 1.7 million in the quarter. The May price increases appear to have stuck without visible churn, so the compression concern is contained for now. The trend is still down year over year, so this is stabilization rather than a turn, but stabilization while volume keeps compounding is the healthier of the two readings.

The mix is also shifting in a favorable direction. Enterprise and government revenue grew 108 percent year over year to $1.8 billion, far faster than consumer at 44 percent, which reduces the company’s dependence on price-sensitive consumer broadband.

Commercial momentum backed that up during the quarter, with new airline agreements including American, Southwest, Virgin Atlantic, Iberia and Aer Lingus, Starlink Mobile partnerships with SoftBank, NTT Docomo and Spark NZ, and FCC approval of the EchoStar spectrum transfer. The counterweight to keep in view is competition. Amazon’s Leo constellation is scaling, and a sustained ARPU decline paired with a credible rival would slow the compounding that the whole model leans on.

Is the AI Segment Real Revenue or a Funded Promise?

The AI segment produced the quarter’s genuine surprise, and also its central caveat. Revenue rose to $2.6 billion, up 247 percent year over year and 213 percent sequentially, well ahead of expectations. The driver was the compute line, where AI solutions and infrastructure revenue jumped to $2.2 billion from less than $0.5 billion in the prior quarter, on $14.1 billion of newly contracted Cloud Services Agreements.

The anchor compute agreements disclosed in the prospectus were Anthropic at about $1.25 billion a month and Google at about $920 million a month, together roughly $26 billion annualized. Both carry 90-day termination clauses, and Google only begins paying in October, so recognition timing was always the variable that would move the segment.

This quarter it landed in the period rather than being deferred to the second half, which many observers had assumed, and the company gave useful detail by splitting advertising from compute and infrastructure. The segment reached its first positive adjusted EBITDA at $1.1 billion.

The caveat is that the profitability is an accounting outcome more than a cash one. The segment still lost $1.3 billion at the operating line, and the swing to positive adjusted EBITDA is explained almost entirely by roughly $1.9 billion of depreciation added back. This was visible at the roadshow, where the deck showed AI at about negative $1.2 billion of adjusted EBITDA against a negative $6.35 billion GAAP operating loss, with depreciation, the core cost of a compute business, set aside. The print follows the same pattern.

The segment looks profitable only once the cost of the infrastructure it is building is set aside, and that cost is enormous. The demand signal is real, helped by the July release of Grok 4.5 and the announced $60 billion acquisition of Cursor, but the economics remain unproven. Revenue is concentrated in a short list of large customers who can terminate on 90 days notice, the Cursor deal raises the level of commitment rather than reducing it, and the segment consumes cash at a scale the next chart makes plain.

Source: SpaceX Q2 2026 results.

The tension in one image. AI turned adjusted EBITDA positive, yet its capital spending in the quarter was more than ten times that figure and dwarfed the rest of the company.

Is Space a Business or a Subsidy Right Now?

Space looks weak on the reported line, and much of that is by design. Revenue grew 29 percent year over year to $962 million, but the operating loss widened to $542 million as spending on the Starship program accelerated, with total segment costs up roughly $389 million from a year earlier.

Launch still flies overwhelmingly for internal Starlink deployment rather than external customers, so reported Space revenue understates the value the segment delivers inside the company, where it builds the constellation that powers the profitable Connectivity business at close to cost.

Operationally, the segment is performing. SpaceX conducted 78 launches and placed 1,041 metric tons into orbit in the first half, remained the world’s leading launch provider, and advanced Starship with a first V3 suborbital flight in May and a July flight that deployed 20 production satellites, relit a Raptor engine in space, and returned with an intact heatshield. The company also booked more than $6 billion of multi-year U.S. government Starshield contracts, which is the clearest near-term source of external Space revenue.

The honest read is that Space is a strategic investment carrying a heavy development cost today, with the larger external opportunity dated to commercial Starship service that the company targets for the second half. It is also the single largest execution dependency in the whole model, since Starship gates the next-generation Starlink satellites and the longer-term ambitions, so any slip in cadence would ripple across the company.

The July flight was a meaningful step toward reducing that risk, but the payoff is still ahead rather than in these numbers.

How Fast Is the Cash Going Out?

Cash is the real constraint and the figure that most deserves attention. Capital spending was $18.4 billion in the quarter, of which about $15.8 billion went to AI infrastructure, against operating cash flow of only $3.5 billion for the first half. Investing outflows for the half reached $34.5 billion. Free cash flow is deeply negative, and the announced Cursor acquisition will add to the bill.

The company can fund this for now. It ended the quarter with about $93.5 billion of cash and $6.5 billion of marketable securities, roughly $100 billion of liquidity, after the IPO proceeds and a $25 billion inaugural bond issued in June at a weighted average rate near 5.9 percent. Total assets more than doubled to $192.8 billion, property, plant and equipment rose to $65.7 billion, and backlog stood at $47.5 billion. The balance sheet strength is real, and it is precisely what the raise was for.

The tension is one of timing rather than solvency. The model depends on the AI build compounding into durable earnings before the cash question moves from theoretical to pressing, and at the current rate of spending that is a matter to track each quarter rather than a distant concern.

The war chest buys years of runway, and the backlog and contracted compute give the spending a purpose, but a company burning cash at this pace has less room for a delay in the payoff than its balance sheet headline implies.

What Does the Share Structure Imply?

One feature of this listing deserves its own treatment, because it shapes how the fundamentals will be priced over the coming months. The float is small, near 4.9 percent of shares after the IPO, and the lockup unwinds on an unusual nine-point schedule rather than a single cliff. The first tranche opens on August 6, two days after this report, and frees roughly 911.5 million shares, more than the entire current public float.

Further tranches of about 7% each follow through the autumn; a larger release comes with the third-quarter report, the remaining initial employee balance frees at the 180-day mark in early December, and Elon Musk, together with major investors, stays locked until June 2027. On that path, the free float rises toward 40 percent of the company by December. Short interest already sits near 32 percent of the float.

Source: SpaceX Q2 2026 results.

The August 6 unlock releases more shares than currently trade, which is why the supply schedule matters more than any single line of the income statement over the next few months.

The demand side runs the other way and is largely spent. Index inclusion forced an estimated $8 to $27 billion of passive buying, with fast entry around July 2 and around August 14, while S&P 500 membership waits until about mid-2027 because it requires seasoning and four profitable quarters. So the mechanical buying that supported the stock early is fading just as the supply arrives.

None of this changes the operating results. It does mean that price behavior over the next several months will be driven as much by who is free to sell as by how the business performs, which is the main reason to judge the company on the fundamentals rather than the tape.

How Does the Entry Valuation Frame the Risk?

Valuation sets the bar the fundamentals have to clear, and SpaceX listed with a very high bar. At the offer, it traded at roughly 92 to 94 times sales, against a mega-cap peer average near 12 times. That is not a small premium to a fast grower; it is a different order of magnitude, and it prices in a decade of the AI and Starship bets working close to plan.

History is not kind to that starting point. In Jay Ritter’s data on listings since 1980, IPOs priced above 40 times sales have averaged a 93.6% first-day gain and then declined about 44.8 percent over the following three years, and among companies with more than $100 million of revenue and a multiple above 40 times sales, 12 of 14 underperformed the market.

Listings that floated less than 5 percent of the company have almost all underperformed over three years as well, which places SpaceX squarely in two cohorts with poor track records.

The counterweight is that base rates are a caution rather than a forecast. The small-float precedents cut both ways: Google never retraced its first-day close, and LinkedIn and ARM recovered strongly after a weak first year; the top decile of IPOs has historically delivered very large returns.

A company with Starlink’s cash generation and a credible AI franchise is not an average listing. What the entry price does is raise the standard of proof, which is why the durability of the AI ramp, and not any single quarter, is what ultimately decides the outcome.

That premium is also reflected in InvestingPro’s valuation models. According to InvestingPro’s Fair Value indicator, the stock appears 7% overvalued, with 11 valuation models pointing to a fair value of around $106 per share.

Source: InvestingPro

Did They Give Investors Anything to Model?

As a first-time reporter, SpaceX provided no forward guidance and no numeric outlook. That leaves the wide range of external estimates for later years unresolved, since the real disagreement among analysts was always about the out-years rather than this quarter. Sell-side revenue forecasts for the coming years already span a very wide range, and nothing in the release narrows it.

The absence of guidance also means the market has little to re-rate against beyond the print itself, so dispersion in views is likely to persist until the company establishes a reporting cadence. The offsetting point is that a first reporter often starts cautiously and adds disclosure over time, and even qualitative segment commentary in future quarters would compress the range. For now, external estimates deserve wide error bars, and the burden of proof sits with each new print rather than with a company forecast.

What Are the Fundamentals Saying?

Stripped of the noise, the quarter says two things at once. The AI revenue became real earlier than expected, which de-risks the top-line thesis and is genuinely bullish. The AI profit did not become real because the reported profitability depends on adding back the depreciation of the infrastructure the company is spending billions to build. And underneath both, Starlink is doing the actual earning, holding price while it scales and tilting toward higher-value demand.

The disconnect between revenue momentum and underlying financial strength is also reflected in InvestingPro’s Financial Health model, which assigns the company a score of 1, placing it in the “Weak Performance” category.

Source: InvestingPro

The bull and bear cases did not change so much as sharpen. The company is executing on the AI build faster than skeptics expected, and it has not yet shown that the build turns into durable earnings rather than durable spending.

That balanced picture is also reflected in InvestingPro’s ProTips, which point to strong growth expectations and healthy liquidity alongside recent share price weakness and a lack of profitability.

Source: InvestingPro

Set against a small float unwinding into a large lockup schedule and an entry valuation near the top of the historical range, the single unresolved question around converting AI investment into sustainable earnings is the one that will define the stock. Everything else in the release is detail around it.

Disclaimer: This article is written for informational purposes only. It is not intended to encourage the purchase of any assets and does not constitute an offer, solicitation, recommendation, or advice to invest. I would like to remind you that all assets are evaluated from multiple perspectives and are highly risky; therefore, any investment decision and the associated risk are the sole responsibility of the investor. Additionally, we do not provide any investment advisory services.